File size: 3,379 Bytes
7500ff7 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 | #!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
👑 QUILLAN-RONIN v5.3.1 / v5.4 — STREAMING INTERACTIVE SOVEREIGN TERMINAL
==================================================================
Run this script to chat directly with Quillan-Ronin v5.3.1 (Canonical) / v5.4 (Frontier).
Automatically selects the best available checkpoint.
"""
import sys
import time
from pathlib import Path
import torch
SCRIPTS_DIR = Path(r"C:\02_QUILLAN\scripts")
CHECKPOINT_DIR = Path(r"C:\02_QUILLAN\checkpoints")
sys.path.insert(0, str(SCRIPTS_DIR))
from quillan_v10_unrolled_sovereign import QuillanConfig, QuillanSovereignUnifiedModel
from sovereign_inference_engine import SovereignTokenizer, SovereignInferenceEngine, SamplingParams
if hasattr(sys.stdout, "reconfigure"):
sys.stdout.reconfigure(encoding="utf-8", errors="replace")
def select_best_checkpoint() -> Path:
"""Auto-select best available checkpoint in priority order."""
candidates = [
CHECKPOINT_DIR / "quillan_frontier_v2_best.pt",
CHECKPOINT_DIR / "quillan_ronin_v531_sovereign_production.pt",
CHECKPOINT_DIR / "checkpoints_sft" / "quillan_super_run_best.pt",
CHECKPOINT_DIR / "checkpoints_sft" / "quillan_sovereign_gold_best.pt",
]
for p in candidates:
if p.exists():
return p
raise FileNotFoundError("No Quillan checkpoint found. Run training first.")
def main():
ckpt_path = select_best_checkpoint()
print("==================================================================", flush=True)
print(" 👑 QUILLAN-RONIN v5.3.1 / v5.4 — INTERACTIVE SOVEREIGN TERMINAL", flush=True)
print(f" Active Checkpoint: {ckpt_path.name}", flush=True)
print("==================================================================\n", flush=True)
device = torch.device("cpu")
tokenizer = SovereignTokenizer("gpt2")
cfg = QuillanConfig(
vocab_size=50257,
hidden_dim=1024,
num_layers=16,
num_heads=32,
num_experts=34,
num_experts_active=4,
max_seq_len=16384,
)
print("[*] Loading Quillan-Ronin weights...", flush=True)
model = QuillanSovereignUnifiedModel(cfg).to(device)
engine = SovereignInferenceEngine.load_from_checkpoint(
model_factory=lambda: model,
checkpoint_path=ckpt_path,
device=device,
)
print("[✓] Model ready! Type 'exit' or 'quit' to end session.\n", flush=True)
history = ""
params = SamplingParams(
max_new_tokens=512,
temperature=0.65,
top_p=0.85,
repetition_penalty=1.20,
frequency_penalty=0.30,
presence_penalty=0.30,
)
while True:
try:
user_input = input("You > ").strip()
if not user_input:
continue
if user_input.lower() in ("exit", "quit"):
print("Exiting session.")
break
prompt = f"<|user|>\n{user_input}\n<|assistant|>\n"
print("Quillan > ", end="", flush=True)
def _stream_cb(token: str):
print(token, end="", flush=True)
engine.generate(prompt=prompt, params=params, stream_callback=_stream_cb)
print("\n")
except KeyboardInterrupt:
print("\nSession interrupted.")
break
if __name__ == "__main__":
main()
|